Badri Adhikari

The paper develops a deep learning method to score protein sequence alignment quality using lDDT scores as indicators.

REALDIST predicts real-valued inter-residue distances for proteins, achieving higher accuracy than existing methods on benchmark datasets.

The paper introduces PDNET, a dataset and framework for predicting protein real-valued distances using deep learning, with open-source tools and web-based training capabilities.

SAXSDom improves domain-domain assembly accuracy using SAXS data and IOHMM, achieving better results on CASP and AIDA datasets.

DEEPCON uses dilated convolutional neural networks with dropout to predict protein contacts, achieving higher precision than existing methods on benchmark datasets.

CONFOLD2 improves contact-driven ab initio protein structure prediction by generating top five models using predicted contacts and a soft square energy function.

DNCON2 improves protein contact prediction using two-level deep convolutional networks, achieving higher precision than existing methods on CASP datasets.